Does chunking affect loading for isel / sel

[ahuang11]

If my files are split across datetimes:

2017050112.nc
2017050212.nc
...
2020050212.nc

Is it better to chunk before loading a selection:

import xarray as xr
ds = xr.tutorial.open_dataset('air_temperature')
ds.chunk({'lat': 1, 'lon': 1}).isel(lat=0, lon=0).load()

Or directly

import xarray as xr
ds = xr.tutorial.open_dataset('air_temperature')
ds.isel(lat=0, lon=0).load()

[andersy005]

Going through Dask's schedulers always adds some overhead.

So, If you don't need to perform out-of-core computation on large arrays, you might be better off without dask/chunking